For the past three years, the artificial intelligence investment narrative has been dominated by a single theme: software. Large language models, chatbots, code generators, and cloud infrastructure have absorbed hundreds of billions of dollars in capital expenditure and captured the imagination of Wall Street. Nvidia's stock has risen more than tenfold. Microsoft, Alphabet, and Meta have committed $680 billion to AI infrastructure in 2026 alone.

But SoftBank Group, the Japanese conglomerate that has arguably placed more bets on transformative technology than any other investor in history, is looking past the screen. Its agreement to acquire ABB's robotics division for $5.375 billion in enterprise value is the boldest statement yet that the next wave of AI investment will not be about what happens on a server. It will be about what happens in a factory, a warehouse, a hospital, and eventually, a home.

What SoftBank Is Buying

ABB's robotics division is not a startup. It is one of the most established names in industrial automation, with a workforce of approximately 7,000 employees and 2024 revenues of $2.3 billion. The division manufactures robotic arms, collaborative robots (cobots), and autonomous mobile robots used in automotive manufacturing, electronics assembly, logistics, and food processing.

What makes the division valuable to SoftBank is not just its hardware, but its installed base. Tens of thousands of ABB robots are operating in factories worldwide, generating real-world data about motion, manipulation, quality control, and material handling. That data is the raw material for training AI models that can operate in physical space, what the industry calls "Physical AI."

The Physical AI Thesis

SoftBank CEO Masayoshi Son has been increasingly vocal about his belief that AI's impact on the physical world will ultimately dwarf its impact on the digital world. The logic is straightforward: the global economy produces roughly $100 trillion in goods and services annually, and the vast majority of that value involves physical processes, manufacturing, logistics, construction, agriculture, healthcare, that have been only marginally touched by automation.

Large language models can write code and summarize documents, but they cannot assemble a car, sort packages in a warehouse, or perform surgery. Physical AI requires a fundamentally different set of capabilities: real-time sensor processing, dexterous manipulation, spatial reasoning, and the ability to operate safely alongside humans. These are problems that ABB's robots have been solving, in narrow ways, for decades.

SoftBank's bet is that combining ABB's hardware expertise and installed base with modern AI models, particularly in areas like computer vision, reinforcement learning, and generative planning, will create robots that are qualitatively different from anything the market has seen before. Not just faster or more precise, but genuinely intelligent: capable of learning new tasks, adapting to unexpected situations, and collaborating with human workers in unstructured environments.

The Portfolio Play

The ABB acquisition does not exist in isolation. SoftBank has spent years assembling a portfolio of robotics-related investments that, taken together, represent the most comprehensive physical AI strategy of any investor in the world. The portfolio includes Berkshire Grey (warehouse automation), AutoStore (storage and retrieval systems), Agile Robots (surgical and industrial robotics), Skild AI (foundation models for robots), and SoftBank Robotics Group (consumer and commercial service robots).

ABB's division fills the most important gap in that portfolio: heavy industrial robotics with global scale. The combined ecosystem gives SoftBank a presence across the entire robotics value chain, from the AI models that make robots intelligent to the hardware platforms that execute in the physical world.

What It Means for Investors

The ABB deal is a signal, not just from SoftBank but from the broader market, that the AI investment cycle is maturing. The "picks and shovels" phase, dominated by GPU manufacturers and cloud providers, is not over. But the application layer is arriving, and physical AI is emerging as the largest addressable market within it.

For investors looking to position ahead of this trend, the public market options remain limited but are growing. Companies like Rockwell Automation, Fanuc, Cognex, and Teradyne offer exposure to industrial automation. Nvidia's Omniverse and Isaac platforms provide the software layer for robotic simulation and training. And the inevitable SoftBank IPO of its robotics portfolio, which Son has hinted at, could create an entirely new category of public investment vehicle.

The deal is expected to close in mid-to-late 2026, subject to regulatory approvals. ABB will receive approximately $5.3 billion in net proceeds and book a pre-tax gain of roughly $2.4 billion, which it plans to return to shareholders.

For ABB, the sale is a strategic simplification. For SoftBank, it is the centerpiece of a thesis that Masayoshi Son believes will define the next decade of technology: the moment when artificial intelligence steps off the screen and into the world.